Diagnosis of gastric inflammation and malignancy in endoscopic biopsies based on Fourier transform infrared spectroscopy.
Fourier transform infrared (FT-IR) (3) spectroscopy can effectively provide chemical information concerning the structure and composition of biological materials at the molecular level; the application of vibrational spectroscopy is therefore expanding (1-5). This noninvasive, convenient, and rapid technique has been applied to the study of various types of healthy and malignant tissues and is a powerful tool in researching the biochemistry of cancer (6-11). The technique also has promising potential for detecting early cancer. Our research group has successfully used FT-IR spectroscopy to diagnose multiple kinds of carcinoma, such as stomach, colon, liver, esophagus, lung, gallbladder, breast, and parotid gland (12-24).
We previously performed FT-IR analysis on a large number of samples obtained at surgery from paired carcinomas and adjacent healthy tissues (diameter, 1-2 cm). The results showed that there were significant differences between the spectra of malignant and corresponding healthy tissues and that a series of discrimination regularities existed (25). In addition, our previous results revealed that FT-IR spectroscopy could detect molecular abnormalities that occurred before the change in morphology seen under the light microscope (25). FT-IR technology thus makes it possible to detect inflammatory and precancerous stages. The FT-IR method has the possibility of developing into a new technique for gastric endoscopic examination. We believe that noninvasive, rapid, accurate, and convenient analysis of gastroscopic tissues can be performed with Fourier-transform mid-infrared spectroscopy if the mid-infrared fiber optics and stomach endoscopy technologies can be combined successfully, but a flexible mid-infrared optical fiber and miniprobe are not yet available. The aim of this study, therefore, was to explore the use of FT-IR spectroscopy, combined with multivariate analysis method, in the classification of gastric endoscopic biopsies.
Materials and Methods
A total of 103 gastric tissue specimens were obtained by the Medical Division of the First Hospital of Xi'an Jiaotong University, China. Informed consent was obtained from each patient before the study. One endoscopic pinch biopsy (diameter, 1-3 mm) was obtained from each patient. According to the results from the pathology review, the studied samples consisted of 19 cases of cancer, 35 cases of chronic atrophic gastritis, 29 cases of chronic superficial gastritis, and 20 samples of healthy stomach tissue.
Because the endoscopic samples are small and it is hard to obtain a high-quality IR spectrum, a modified attenuated total reflectance (ATR) accessory was linked to a WQD500 FT-IR spectrometer (Beijing No. 2 Optical Instrument Factory). The modified ATR accessory was constructed with a ZnSe crystal. The improvement in the ATR accessory involves reducing the number of internal reflectances to five, which can decrease the attenuation of light energy and provide higher light throughput, thus increasing the signal-to-noise ratio. The FT-IR spectrometer was equipped with a mercury cadmium telluride detector cooled by liquid nitrogen, which has higher stability and sensitivity than the traditional deuterated triglycinate sulfate detector.
The fresh surgically resected specimens were immediately noninvasively analyzed with the mobile WQD-500 FT-IR spectrometer. Each sample was placed on the ATR accessory for analysis. To collect the data for each spectrum, we performed 32 coadded scans at a resolution of 4 cm-1, with a typical range from 800 to 4000 cm-1. It took -1-2 min to obtain the spectrum noninvasively. For comparison, after the spectra for the samples were recorded, the fixed samples were stored in liquid nitrogen and sent for histologic examination for comparison with the spectral analysis. Each IR result was compared with a biopsy from the same tissue sample.
MULTIVARIATE STATISTICAL ANALYSIS
After FT-IR spectra were collected, and in parallel, a pathology-based diagnosis was made using the routine method of hematoxylin/eosin staining for the property of interest. In the second step, we statistically analyzed the obtained FT-IR spectra to determine differences and convert the spectroscopic data into clinically useful information.
We used OMNIC5.0 as the data-processing software. The five-point moving average smoothing method was adopted for each spectrum to reduce the random noise in the data. To achieve more accurate differences in the ratio of peak intensities, we first performed baseline corrections of the spectra. For each spectral feature, such as the relative intensities of peaks A and B, we performed baseline corrections between the peaks being compared to sharpen the spectral differences among different kinds of tissues. For the purpose of multiple group classifications, we performed a supervised linear discriminant analysis (LDA), using the SPSS 10.0 statistical package (26) for the LDA algorithm.
For a supervised LDA, the class identity was known and used in the calculation. The LDA algorithm was derived from an equation (27):
Z = [w.sub.1][x.sub.1] + [w.sub.2] [X.sub.2] + [w.sub.3][x.sub.3] + ... + [w.sub.i][x.sub.i]
where Z is the discriminant score, and [w.sub.1] is the discriminant weight for the ith independent variable [x.sub.i]. In this study, the four classifications of gastric endoscopic biopsies were identified by LDA algorithms with seven independent variables, which were the intensity ratios of ~1640 [cm.sup.-1] (peak A) vs ~1550[cm.sup.-1] (peak B), ~1460 cm-1 (peak C) vs ~1400[cm.sup.-1] (peak D), ~1310[cm.sup.-1] (peak E) vs ~1240[cm.sup.-1] (peak F), and ~1160[cm.sup.-1] (peak G) vs ~1120[cm.sup.-1] (peak H), and the peak positions of the amide I band plus water (peak A), amide II band (peak B), and ~1310[cm.sup.-1] band (peak E), respectively. Because the number of samples for each group was not large enough, all spectra were split into a training set to train the LDA models, and a leave-one-out cross-validation was used to assess the quality of the discriminant model. The class assignment of any given spectrum was derived by computing its distance from all class centroids and allotting it to the class whose centroid was nearest.
In the endoscopic biopsies, the majority of spectral peaks were in the 1000-1800[cm.sup.-1] region. The mean FT-IR absorption spectra for the gastric endoscopic biopsies in the regions 1800-1480[cm.sup.-1] and 1480-1000[cm.sup.-1] are shown in Fig. 1. Similar to previous results, there was distinctive absorption at peaks A through H, which were selected as the seven independent variables for multivariate discriminant analysis.
[FIGURE 1 OMITTED]
The statistical results for the intensity ratios [mean (SD)] for peaks A/B, C/D, E/F, and G/H and the mean (SD) peak positions for A, B, and E in the FT-IR spectra for different types of gastric endoscopic samples are listed in Table 1. We observed that the change in the regularities of the spectral characteristics agreed with the type of disease. Therefore, all of the characteristics listed above needed to be considered simultaneously for the classification of gastric tissues obtained from endoscopic examination.
In this study, we first used three discriminant functions in the analysis and used three-dimensional scores for each sample for classification. The classification results for all four groups of gastric specimens in the training set and the cross-validation study are shown in Table 2. The results based on the FT-IR spectra with the LDA analysis, with reference to the histologic diagnoses, are shown in Table 3. The LDA-supervised classification was very promising, with 84% overall accuracy for the training set. For the cross-validation, each case was classified by the first three discriminant functions derived from all cases other than that individual case, with an overall accuracy of 77% correctly classified for the cross-validated grouped cases. The sensitivities for FT-IR spectroscopic classification of gastric cancer, chronic atrophic gastritis, superficial gastritis, and healthy tissue were 74%, 66%, 90%, and 90%, respectively.
The spectral characteristics of peaks A-H in the FT-IR spectra differed among healthy, chronic superficial gastritis, atrophic gastritis, and cancerous gastric specimens. Significant shifts of the peak frequencies and changes in the intensity ratios occurred, as shown in Table 1 and Fig. 1. The spectral characteristics of the small gastric biopsies were similar to those of block gastric tissues, which were obtained during surgery and measured in our previous studies. The spectral features of malignant and healthy gastric tissues were in good agreement with the criteria established in our previous work (21).
These spectral features are related to the changes in structure and composition of biological molecules such as proteins, nucleic acids, and fats. As shown in Fig. 1, the shape of spectra for chronic atrophic gastritis exhibited some of the same characteristic bands as cancer. Chronic superficial gastritis is a milder inflammation, and the shape of the spectrum for this condition was similar to that of healthy gastric tissue. The prominent band at ~1640[cm.sup.-1] belongs to the amide I band of protein and the H-O-H deformation vibration of water; the -1550 cm-1 absorption peak arises from N-H bending and C-N stretching (amide II band) in proteins. In the spectra of gastric cancer biopsies, the ratios of the intensities of peak A to peak B were higher in the malignant tissue because the intensity of peak A was stronger as a result of the higher water content in cancerous tissue. In addition, the strength of the amide II bands gradually increased from chronic atrophic gastritis to superficial gastritis to healthy tissue, as shown in Fig. 1.
The intensity of the -1400[cm.sup.-1] peak was stronger than that of the ~1460[cm.sup.-1] peak in the spectra of cancerous samples. For chronic atrophic gastritis, the decrease in intensity near 1460[cm.sup.-1] was not significant, i.e., the intensity of the band at ~1460[cm.sup.-1] became slightly less than, or approximately equal to, that of the peak at 1400[cm.sup.-1]. In the spectra of superficial gastritis samples, the intensity of the peak at 1460[cm.sup.-1] was slightly stronger than that of the peak at 1400[cm.sup.-1], and the intensity of the peak at 1460[cm.sup.-1] was stronger than that of the peak at 1400[cm.sup.-1] in healthy tissue.
In the spectra for cancer, the intensity of the absorption peak at ~1310[cm.sup.-1] increased, and the peak position shifted to a lower wavenumber. Compared with the spectra of the malignant tissues, the absorption peak near 1310[cm.sup.-1] was weaker in the spectra of atrophic gastritis samples. Similar to the gastric cancer tissues, the position of peak E often shifted to a lower wavenumber, which was different from chronic superficial gastritis. In healthy tissues, the peak at ~1240[cm.sup.-1] was stronger, and the band near 1318[cm.sup.-1] became weak and sometimes disappeared. In addition, the position of this band often shifted to a higher wavenumber.
The intensity of the band at ~1160[cm.sup.-1] was often less than that of the band at ~1120[cm.sup.-1] in the spectra of stomach cancer samples. In the healthy gastric tissues, the intensity of the peak near 1160[cm.sup.-1] increased and was often stronger than that of band at ~1120[cm.sup.-1].
Our results indicate that FT-IR spectroscopy, combined with appropriate pattern recognition algorithms, can distinguish benign from malignant disease for endoscopic gastric biopsies. There are several advantages of FT-IR spectroscopy: it is inexpensive, less time-consuming, does not require special sample preparation or any biochemical reagents, and uses only small amounts of sample, leaving sufficient material for other clinical tests. Moreover, it is a computer-operated system, which helps standardize interpretation of results. Real-time FT-IR spectroscopy of biopsy samples taken at the time of endoscopy can provide accurate, rapid diagnostic differentiation between healthy tissue, superficial and atrophic inflammation, and cancer, using a technique that is currently available. If a miniprobe can be developed for IR detection that can pass through the endoscopic biopsy channel and be coupled with fiber optics, clinicians would have a noninvasive real-time technique for directed instead of random biopsy analysis.
We thank the National Natural Science Foundation of China (Grant 30371604), the State Key Project for Fundamental Research of China (Grant 2002CCA01900), and the Doctoral Project of the High Education Ministry of China for supporting this work. We gratefully acknowledge Beijing No. 2 Optical Instrument Factory (Beijing, China) for providing the FT-IR spectrometer and for excellent technical assistance.
(1.) Graham J, Leslie B, Antony W, Marinos E. The use of infrared spectrophotometry for measuring body water spaces. Clin Chem 1999;45:1077-81.
(2.) Franck P, Sallerin JL, Schroeder H, Gelot MA, Nabet P. Rapid determination of fecal fat by Fourier transform infrared analysis (FTIR) with partial least-squares regression and an attenuated total reflectance accessory. Clin Chem 1996;42:2015-20.
(3.) Rahman J, Hewlett A, Taylor DR, Xiao SY, Wu JG, Soloway RD. Reproducible, rapid quantitation of hepatic fibrosis in human liver biopsies using Fourier transform infrared (FT-IR) spectroscopy. Gastroenterology 2002;122(Suppl):A303.
(4.) Kruse-Jarres JD, Janatsch G, Gless U, Marbach R, Heise HM. Glucose and other constituents of blood determined by ATR-FTIR-spectroscopy. Clin Chem 1990;36:401-2.
(5.) Christopher VE, Arnold MA. Near-infrared spectroscopy for measuring urea in hemodialysis fluids. Clin Chem 2001;47:1279-86.
(6.) Rigas B, Morgello S, Goldman IS, Wong PT. Human colorectal cancers display abnormal Fourier-transform infrared spectra. Proc Natl Acad Sci U S A 1990;87:8140-4.
(7.) Koenig F, Larne R, Enquist H, Mcgovern FJ, Schomacker KT, Kollias N, et al. Spectroscopic measurement of diffuse reflectance for enhanced detection of bladder carcinoma. Urology 1998;51: 342-5.
(8.) Wang HP, Wang HC, Huang YJ. Microscopic FTIR studies of lung cancer cells in pleural fluid. Sci Total Environ 1997;204:283-7.
(9.) Huleihel M, Salman A, Erukhimovitch V, Ramesh J, Hammody Z, Mordechai S. Novel spectral method for the study of viral carcinogenesis in vitro. J. Biochem Biophys Methods 2002;50:111-21.
(10.) Fung Kee Fung M, Senterman M, Eid P, Faught W, Mikhael NZ, Wong PT. Comparison of Fourier-transform infrared spectroscopic screening of exfoliated cervical cells with standard Papanicolaou screening. Gynecol Oncol 1997;66:10-5.
(11.) Argov S, Ramesh J, Salman A, Sinelnikov I, Goldstein J, Guterman H, et al. Diagnostic potential of Fourier-transform infrared microspectroscopy and advanced computational methods in colon cancer patients. J Biomed Opt 2002;7:248-54.
(12.) Wang JS, Shi JS, Xu YZ, Duan XY, Zhang L, Wang J, et al. FT-IR spectroscopic analysis of normal and cancerous tissues of esophagus. World J Gastroenterol 2003;9:1897-9.
(13.) Zhao DW, Zhou XS, Soloway RD, Peng Q, Wu JG, Liang Z, et al. Infrared spectroscopic features of colon cancer cells differ from those of normal colonic cells from the same patient. Gastroenterology 1996;110(Suppl):A620.
(14.) Li WH, Peng Q, Soloway RD, Peng Q, Wu JG, Pena LDO, et al. Surface proteins of cancer and normal epithelium from the same patients differ as assessed by their primary and second structure determined by Fourier transform infrared (FT-IR) spectroscopy. Gastroenterology 1997;112(Suppl):A604.
(15.) Wang JS, Shi JS, Xu YZ, Duan XY, Zhang L, Zhu AJ, et al. FT-IR spectroscopic analysis of gallbladder tissues. Chin J Hepatobiliary Surg 2003;9:657-60.
(16.) Ren Y, Xu YZ, Wang J, Zhang YF, Wang F, Shi JS, et al. FTIR spectroscopic and statistical studies on the lung tissues. Spectrosc Spect Anal 2003;23:681-4.
(17.) Yang LM, Xu Z, Zhang YF, Xu YZ, Zhou S, Zhang NW, et al. The application of mid-IR fiber optics in diagnosis of colon cancer in vivo and in situ. Spectrosc Spect Anal 2003;23:883-4.
(18.) Xu YZ, Zhang YF, Yang LM, Peng X, Sun KH, Li QB, et al. In vivo and rapid detection of the malignancy tumor of oral tissues using Mid-FTIR fiber optics spectroscopy. Chem J Chin Uni 2004;25: 348-50.
(19.) Wu JG, Xu YZ, Sun CW, Soloway RD, Xu DF, Wu QG, et al. Distinguishing malignant from normal oral tissues using FTIR fiber-optic techniques. Biopolymers 2001;62:185-92.
(20.) Wang F, Ling XF, Yang ZL, Xu Z, Ren Y, Li WH, et al. FTIR spectroscopic study of normal and malignant tissues of rectum. Spectrosc Spect Anal 2003;23:498-501.
(21.) Xu YZ, Soloway RD, Ling XF, Zhi X, Weng SF, Wu QG, et al, Xu DF. Fourier transform infrared (FT-IR) mid-IR spectroscopy separates normal and malignant tissue from the colon and stomach. Gastroenterology 2000;118:A6438.
(22.) Sun XJ, Su YL, Soloway RD, Zhang L, Wang JS, Ren Y, et al. Rapid, intraoperative detection of malignancy using attenuated total reflectance (ATR) and mobile Fourier transform infrared (FT-IR) spectroscopy. Gastroenterology 2003;124(Suppl):A420-1.
(23.) Peng Q, Xu YZ, Li WH, Zhou XS, Wu JG. FTIR study on the normal and tumor gastrointestinal tissues. Spectrosc Spect Anal 1998; 18:528-31.
(24.) Sun CW, Xu YZ, Sun KH, Wu QG, Li WH, Xu ZH, et al. A study of the diagnosis of salivary gland tumors by means of mid infrared optical fiber technique. Spectrosc Spect Anal 1996;16:22-5.
(25.) Zhang L, Sun KH, Soloway RD, Ling XF, Xu YZ, Wu QG, et al. Intraoperative Fourier transform infrared spectroscopy can guide individual resections in patients with gastric cancer. Gastroenterology 2004;126:A626.
(26.) Liu RX, Kuang J, Gong Q, Hou XL. Principal component regression analysis with SPSS. Comput Methods Programs Biomed 2003; 71:141-7.
(27.) Fujioka N, Morimoto Y, Arai T, Kikuchi M. Discrimination between normal and malignant human gastric tissues by Fourier transform infrared spectroscopy. Cancer Detect Prev 2004;28:32-6.
QING-BO LI,  XUE-JUN SUN,  YI-ZHUANG XU,  LI-MIN YANG,  YUAN-FU ZHANG,  SHI-FU WENG,  DING-SEN SHI,  and JIN-GUANG WU  *
 The State Key Laboratory of Rare Earth Materials Chemistry and Applications, College of Chemistry and Molecular Engineering, Peking University, Beijing, Peoples Republic of China.
 Department of General Surgery, First Hospital of Xi'an Jiaotong University, Xi'an, Shanxi Province, Peoples Republic of China.
* Address correspondence to this author at: The State Key Laboratory of Rare Earth Materials Chemistry and Applications, College of Chemistry and Molecular Engineering, Peking University, Beijing 100871, Peoples Republic of China. Fax 86-10-62751708; e-mail firstname.lastname@example.org.
Received June 1, 2004; accepted December 10, 2004. Previously published online at DOI: 10.1373/clinchem.2004.037986
 Nonstandard abbreviations: FT-IR, Fourier transform infrared; ATR, attenuated total reflectance; and LDA, linear discriminant analysis.
Table 1. Mean (SD) peak frequencies and intensity ratios for healthy gastric tissue, chronic superficial gastritis, chronic atrophic gastritis, and gastric cancer. Chronic superficial Healthy tissue gastritis (n = 20) (n = 29) Peak frequency, [cm.sup.-1] A 1646 (2) 1642 (3) B 1553 (2) 1546 (2) E 1317 (6) 1316 (5) Intensity ratio A/B 1.65 (0.19) 2.12 (0.26) C/D 1.10 (0.04) 1.40 (0.67) E/F 0.67 (0.09) 0.34 (0.21) G/H 1.22 (0.25) 0.91 (0.06) Chronic atrophic gastritis Cancer (n = 35) (n = 19) Peak frequency, [cm.sup.-1] A 1640 (3) 1641 (2) B 1547 (2) 1549 (4) E 1306 (8) 1313 (3) Intensity ratio A/B 2.33 (0.31) 2.48 (0.30) C/D 0.91 (0.34) 0.86 (0.10) E/F 0.52 (0.38) 0.79 (0.32) G/H 0.85 (0.12) 0.89 (0.13) Table 2. Comparison of the FT-IR results with histologic examination. Histologic examination Chronic atrophic FT-IR results Cancer gastritis Training set Cancer 16 7 Chronic atrophic gastritis 1 25 Chronic superficial gastritis 2 3 Healthy 0 0 Cross-validation Cancer 14 9 Chronic atrophic gastritis 2 23 Chronic superficial gastritis 3 3 Healthy 0 0 Histologic examination Chronic superficial FT-IR results gastritis Healthy Training set Cancer 1 2 Chronic atrophic gastritis 0 0 Chronic superficial gastritis 28 0 Healthy 0 18 Cross-validation Cancer 1 2 Chronic atrophic gastritis 1 0 Chronic superficial gastritis 26 0 Healthy 1 18 Table 3. Results of statistical analysis of detection of endoscopic gastric samples by FT-IR spectroscopy. Sensitivity, % Specificity, % Training set Cancer 84 88 Chronic atrophic gastritis 71 98 Chronic superficial gastritis 97 93 Healthy 90 100 Cross-validation Cancer 74 86 Chronic atrophic gastritis 66 96 Chronic superficial gastritis 90 92 Healthy 90 99 Predictive Predictive value of a value of a positive test, % negative test, % Training set Cancer 62 96 Chronic atrophic gastritis 96 87 Chronic superficial gastritis 85 99 Healthy 100 98 Cross-validation Cancer 54 94 Chronic atrophic gastritis 88 84 Chronic superficial gastritis 81 96 Healthy 95 98
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Cancer Diagnostics|
|Author:||Li, Qing-Bo; Sun, Xue-Jun; Xu, Yi-Zhuang; Yang, Li-Min; Zhang, Yuan-Fu; Weng, Shi-Fu; Shi, Ding-Sen;|
|Date:||Feb 1, 2005|
|Previous Article:||Simultaneous measurement of [[beta]-amyloid.sub.(1-42)], total tau, and phosphorylated tau ([Thr.sup.181)] in cerebrospinal fluid by the xMAP...|
|Next Article:||The novel apolipoprotein A5 is present in human serum, is associated with VLDL, HDL, and chylomicrons, and circulates at very low concentrations...|